
Documento original
GPT Image 2 · en
Scene:
A square cinematic studio photograph of a Mars habitat diorama cube under warm directional light. The cube rests on a minimal black base, with a rust-red dust haze and a dark neutral background. Use a sharp three-quarter view that shows the top habitat surface and multiple cutaway sides with clean, physically convincing edges.
Subject:
The top face shows a tiny modular Mars research outpost: pressurized domes, solar arrays, rover tracks, dust-covered equipment, regolith berms, and a small greenhouse bubble. The vertical cut faces expose layered red regolith, basalt fragments, buried ice lenses, cable trenches, habitat foundation pylons, a sealed lava tube chamber, and older impact breccia. Include fine dust, compacted tire marks, small rocks, and realistic scale cues.
Important details:
The cube should feel like a future planetary science exhibit. All infrastructure must connect logically: cables should run from the surface into trenches, foundations should enter the regolith, and the greenhouse should sit on a stable pad. Subsurface ice must appear as pale translucent seams, not blue fantasy crystal. The lava tube should be a dark hollow space below the habitat, with rough basalt walls and no magical glow.
Use case:
Hyperreal Diorama Cube single study 4/8, testing a Mars habitat surface with regolith structure, buried ice, infrastructure, and a lava-tube void.
Constraints:
No readable text, no watermark, no logos, preserve exact 1:1 aspect ratio, no real space agency marks, no astronauts posing, no flags, no floating labels, no oversized rover, no glowing alien artifacts, no disconnected underground cables, no blurry side strata.Cómo usar
- Copia el prompt original completo, incluidos los parámetros.
- Adapta el tema, el texto y la composición a tus necesidades.
- Usa una herramienta externa compatible; comprueba el acceso y el resultado.
Antes de empezar
- Comprueba si necesita una imagen de referencia. La compatibilidad no demuestra qué modelo generó el ejemplo.
- Revisa nombres, cifras y textos; el resultado puede variar entre versiones del modelo.





